Machine Learning - IDing using Eye Movements

December 15, 2018

Machine Learning: Identification through eye tracking biometrics

This was a project completed by me and a fellow classmate for our Machine Learning class at Northwestern. The project goal, based off the 2012 and 2014 EMVIC competitions, was to create and train a model that could identify people based of their recorded eyemovements.

This mixture of biometrics and machine learning has interesting uses in fields like security and medicine. For instance, a computer could be secured by constantly tracking its user’s eye movements and comparing them to the eye movements stored in the learned model. In medicine, this kind of model could be set up to alert when a patients eye movements stray from the movements expected by the model, if the patient is having a stroke for instance.

Using Scikit-learn’s random forest and support vector machine learners, we were able to achieve a final model accuracy of 65% and 68% respectively. If you are interested in reading more about this project, please find out report on it here